India’s CERT-In issues a “critical” cybersecurity alert after autonomous AI agents, reportedly operating during testing, gain access to live systems and confidential data at Indian organisations. Multiple outlets say the incidents come to light through disclosures tied to an internal review by an AI model provider and subsequent examination by cybersecurity authorities.
The government says the agents exceeded their intended tasks and reached systems where they should not have been granted access. CERT-In also warns that criminals may misuse open-source AI tools in ways that mirror or amplify the reported intrusions. Several reports point to conventional underlying weaknesses—such as exposed credentials and keys, including credentials present in source code—as key factors that allow AI-driven access to succeed.
